Aaron Schwartz

Aaron Schwartz

Assistant Professor, Department of Medical Ethics and Health Policy and Department of Medicine, University of Pennsylvania Perelman School of Medicine

Dissertation Title:"Measuring Health Care Quality and Value: Theory and Empirics"Imperfect information is a pervasive feature of health care markets. Therefore, measuring the quality and value of health care services may inform efforts to improve health care delivery.  This dissertation explores several applications of performance measurement in health care: describing national practice patterns, evaluating the effects of payment reforms, and contributing to policies that reward providers for measured performance.

Chapter one describes the use of low-value services in fee-for-service Medicare.  Drawing from evidence-based lists of services that provide minimal clinical benefit, I develop 26 claims-based measures of low-value services.  Applying these measures to Medicare claims, I demonstrate that 42% of beneficiaries received at least one of these services in a year, which constituted 2.7 % of overall annual spending.  When more specific and less sensitive versions of the measures were used, I detected low-value service use for 25% of beneficiaries, constituting 0.6% of overall spending.  In adjusted analyses, spending on low-value services was substantial even in regions at the 5th percentile of the regional distribution of low-value spending.  Adjusted regional use was positively correlated among five of six categories of low-value services. These findings are consistent with the view that wasteful practices are pervasive in the US health care system.  The results also suggest that the performance of claims-based measures in supporting policies to reduce overuse may depend heavily on how the measures are defined.

Chapter two examines the role of provider organizations in influencing the delivery of low-value services.  In Part I of this chapter, I assess whether provider organizations exhibit distinct profiles of low-value service use in fee-for-service Medicare.  In one sample of 3,137 large provider organizations and another sample of 250 provider organizations that entered the Medicare Pioneer Accountable Care Organization (ACO) Program or the Medicare Shared Savings Program, I demonstrate that provider organizations’ use of low-value services exhibits considerable variation, substantial persistence over time, and modest consistency across service types.  In Part II of this chapter, I evaluate the effects of the Pioneer ACO Program on the use of low-value services. In a difference-in-differences analysis, I compare the use of low-value services between beneficiaries attributed to Pioneer ACOs and beneficiaries attributed to other providers, before (2009-2011) vs. after (2012) Pioneer ACO contracts began. During its first year, the Pioneer ACO program was associated with modest reductions in low-value services, with greater reductions for organizations that had provided more low-value services. The findings in this chapter suggest that provider organizations can influence the use of low-value services by affiliated physicians, and that organization-level incentives can reduce low-value practices.

Chapter three analyzes the economic properties of performance measures used in both health care and education policy. Because observable outcomes constitute a noisy signal of performance in these settings, shrinkage estimators are often used to improve measurement accuracy. I demonstrate that these improvements in accuracy come at the cost of reducing a measure’s responsiveness to agent behavior, thereby diluting incentives for performance improvement.  In a model of consumers sorting between agents, I show that welfare depends on two components: (1) accuracy of performance signals, which promotes efficient consumer sorting, and (2) incentives for performance improvement, which promote efficient agent effort.  Using Monte Carlo simulation, I evaluate the accuracy and incentive properties of various techniques for estimating hospital performance in heart attack mortality.  Shrinkage estimators entail substantial incentive distortions, particularly for smaller hospitals, which experience an approximate 50-70% “tax” on improvement.  Several estimation techniques, including the methods currently used by Medicare, are dominated on the basis of both accuracy and incentive criteria.  I discuss various policy alternatives to shrinkage estimation, such as increasing the timespan of measuring performance.


Graduation Year